The present invention relates to semiconductor processing technology and, in particular, concerns a device and a fabrication process whereby a Magnetoresistive Random Access Memory (MRAM) structure can be used as a selectively programmable antifuse device.
Since the introduction of the digital computer, electronic storage devices have been a vital resource for the retention of binary data. Conventional semiconductor electronic storage devices incorporate capacitor type structures, which are referred to as Dynamic Random Access Memory (DRAM) that temporarily store binary data based on the charged state of a capacitor. This type of semiconductor Random Access Memory (RAM) requires a continuous supply of power and a periodic charge refresh to maintain a particular defined logic-state. As a result, semiconductor RAM is considered volatile memory due to the fact that data can be lost with the loss of power.
For many applications, it may be desirable to replace traditional volatile memory with an improved solid-state non-volatile memory device. This need has fueled research and development in the area of non-volatile memory storage devices while still maintaining a high-density fabrication process and technique. The increased demand for a more sophisticated, efficient, and non-volatile data retention technique has driven the development of Magnetoresistive Random Access Memory (MRAM) devices.
MRAM is a developing technology that offers the advantages of non-volatility and high-density fabrication. MRAM structures employ the spin property of electrons within layers of magnetic material to read the memory storage logic states. Binary logic states typically take advantage of a resistance differential to distinguish between “on” and “off” states. Resistance is a measure of the inability of an electrical current to flow through a specific material, whereas current is the actual flow of charge carriers through a material. If a material has a high resistance, then the ability of electrons to flow through the material is inhibited. Conversely, a low resistive material tends to allow a higher degree of current to flow.
MRAM structures take advantage of this resistivity concept by manipulating the alignments of spin states within multiple layers of material to increase or decrease the resistance of a material. Magnetic Tunnel Junctions (MTJ), for example, are electronic structures that exhibit a high resistance across the tunneling dielectric structure. MTJ based MRAM devices incorporate at least two thin layers of magnetic material separated by an insulating tunnel barrier.
The magnetic and insulating layers are fabricated on the substrate, The MRAM magnetic layers consist of a magnetic pinned layer and a magnetic free layer. The selective programmability of the sense layer enables the MRAM structure to act as a logic state device, which stores binary data as directions of net magnetization vectors in the metallic MTJ layers. Current flow through two orthogonal conductive traces induces a magnetic moment in a parallel or anti-parallel configuration between the pinned layer and the sense layer. Alternatively, current flow through the MTJ layers, can change the orientation of the magnetization in the free layer through the spin-torque-transfer effect.
MRAM structures employ the nature of spin dependent variable resistance when used to define logic states wherein the high and low resistivity states represent a logical “1”, or “0,” respectively. The corresponding parallel and antiparallel magnetization states reflect and represent two different resistances. The overall MTJ resistance has minimum and maximum values when the net magnetization vectors of the two individual magnetic layers point in substantially the same (parallel) and opposite (antiparallel) directions, respectively. When these materials are layered in a particular fashion, they exhibit a variable vertical electrical resistance depending on the magnetization state of the individual layers.
Due to the physical arrangements of MTJ layers, MRAM structures may be fabricated using similar integration techniques as conventional DRAM. In addition, integrated circuits often implement permanent programmable electrical connections between circuit nodes. They are useful to repair defective elements on the chip and to customize the chip for different applications. One type of permanent programmable electrical connections is referred to as an antifuse link. Field Programmable Gate Array (FPGA) devices may employ antifuse structures for their high-density fabrication ability and efficient ease of programmability. Application Specific Integrated Circuits (ASICs) may make use of antifuses for effective system logic implementation on a single chip.
A fuse is often used in an electrical circuit to inhibit the excessive flow of electrons by creating a permanent open circuit at the fuse. When a fuse is “blown,” the permanently open circuit stops the flow of charge carriers through the electrical circuit at the fuse. In contrast, an antifuse is often used in an electrical circuit to allow an open flow of charge carriers by creating a permanent short circuit at the fuse. When an antifuse is “blown,” a permanent short is created at the antifuse.
Conventional antifuse devices in a DRAM array may be fabricated with a structure similar to that of a capacitor, i.e., a dielectric layer separated by two conductive electrical terminals. Initially, an antifuse device is fabricated in the “off” state, which presents a high resistance between antifuse terminals. Conversely, the “on” state represents a lower resistive connection between antifuse terminals. To program an antifuse “on” state, a large programming voltage, usually between 7 and 9 volts, may be applied across the antifuse terminals, which causes breakdown of the interposed dielectric and forms a short circuit between the antifuse terminals. However, the short circuit produced by high voltage can be unreliable. Accordingly, alternative methods, such as laser fusing is generally employed to program antifuses in DRAM circuits. Though more reliable, such processes are cumbersome, particularly where they depend upon the use of different materials on structures from adjacent cell capacitors, which may require additional processing steps during fabrication. An even larger voltage could be used to better break down the dielectric material, but such a voltage is more likely to cause damage to neighboring circuitry.
Hence, the use of conventional antifuse devices in MRAM types of devices may be expensive in terms of production costs and in terms of loss of available space for other circuit components. Additionally, the use of conventional antifuse devices may potentially cause damage to neighboring circuit components due to the relatively high programming voltage needed to change the antifuse from a non-conducting state to a conducting state. Thus, there exists a need for an improved method of manufacturing antifuse devices that do not consume as much substrate space and do not add additional processing steps to the process of fabricating integrated circuitry. More particularly, there is a need for an antifuse device that is better suited for fabrication with MRAM type devices that may be programmed in a safer manner.